White matter biomarker for predicting de novo Parkinson's disease using tract-based spatial statistics: a machine...

Area of Science:

• Neuroimaging
• Neurology
• Biomarker Discovery

Background:

• Parkinson's disease (PD) is a chronic, neurodegenerative disorder.
• Cerebral white matter (WM) lesions are associated with PD.
• Early detection of WM alterations can aid in timely intervention.

Purpose of the Study:

• To investigate microstructural alterations in white matter (WM) in early Parkinson's disease (PD).
• To develop a predictive model for early PD diagnosis using WM lesions.
• To assess the impact of integrating clinical factors into the predictive model.

Main Methods:

• Utilized data from the Parkinson's Progression Markers Initiative (PPMI) database (152 PD patients, 75 controls).
• Performed whole-brain voxel analysis of WM using tract-based spatial statistics (TBSS).
• Employed least absolute shrinkage and selection operator (LASSO) regression and random forest (RF) algorithms for classification, with 5-fold cross-validation.

Main Results:

• PD patients showed significantly decreased fractional anisotropy (FA) values, indicating widespread WM lesions.
• The WM lesion-based PD prediction model achieved an accuracy of 0.783 and AUC of 0.831 in the test set.
• Integrating clinical factors improved the model's test set performance to 0.804 accuracy and 0.844 AUC.

Conclusions:

• Quantitative analysis of WM lesions on FA images via TBSS serves as a neuroimaging biomarker for early PD diagnosis.
• The predictive model demonstrates potential for individual-level early PD detection.
• Integration with clinical variables enhances the predictive power of the neuroimaging biomarker.